Interaction network analysis in postgwas.

<p>Part (A) shows the complete network of genes derived from the human height GWAS dataset <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0071775#pone.0071775-LangoAllen2" target="_blank">[38]</a> using a p-value cutoff at 1×10⁻⁶, generated by application of the <i>postgwas()</i> summary function to the dataset without further customization parameters. Appearance of the network can be modified by using a custom (drag and drop) vertex layout or deactivated edge labels. The edges of this network are formed based on common REACTOME pathway membership, (optionally) labeled by the type of interaction (here: shared pathway name) and weighted by the combined association strength of participating genes. Vertex sizes (and optionally transparency) correspond to the GWAS association p-value. Under consideration of these weights, application of a minimum cut-edge graph partitioning algorithm leads to a decomposition of the global graph into functional subunits with preferential accumulation of well-associated genes within modules. Part (B) shows the first extracted module, exhibiting the strongest evidence for accumulation of low GWAS p-values. This accumulation is reflected by a module score listed in the legend (right box, a lower score corresponds to higher evidence). The major biological functions are identified for each module by GO-term over-representation analysis. The top three over-represented terms are listed in a colorized legend together with the module score. Vertices within the module are colorized according to their membership in over-represented GO terms. Part (C) demonstrates a network analysis for multiple datasets. The module shown has been extracted from a network of GO term similarity between genes from two distinct synthetically generated GWAS datasets. Each dataset corresponds to a vertex shape (squares and circles). For genes occurring in both datasets, vertex shapes are plotted on top of each other in order of their p-value (e.g. SLC38A4). Their label is printed boldface and italic. When a single SNP is annotated to multiple genes (e.g. residing in a larger LD block), these genes are labeled with a cross as shown for three solute carrier genes residing in the same block and having similar molecular functions. Such modules need careful interpretation.